Blog

Startups set to scale operations face a multitude of challenges at every stage. These include leveraging the right technology and new-age solutions, protecting customer data, ensuring data security, opting for a scalable tech stack, and staying compliant. Scaling too quickly and without the right IT infrastructure can overwhelm founders and companies.

Oracle, in association with YourStory Media, recently conducted a roundtable discussion on ‘Boosting Digital Innovation with Data, Infrastructure and AI solutions: from Startup to Scaleup’ to discuss the challenges that startups face and how the right solutions can unlock sustainable growth, security, and scalability.

Moderator Rishabh Mansur, Head – Community, YourStory Media, was joined by Tanika Gupta, Director, Data Science, Sigmoid; Sathvik Vishwanath, Co-founder and CEO, Unocoin; Nagarjuna Kandimala, Associate Director- Architecture & Cloud CoE Lead, Carelon Global Solutions; Diwakar Chittora, Founder and CEO, Intellipaat; Somit Pangtey, Head of Engineering, ClickPost; Sarat Buddhiraju, Chief Architect, Bigbasket; Ravi Suhag, VP Engineering, Pixxel; Praveen DS, Co-founder and CTO, Lemnisk; Shubham Shukla, VP Mobile Engineering, Flam; Rajath Kedilaya, VP, Technology and Product, DrinkPrime; Sai Movva, Director of Engineering & Staff Software Engineer, Scaler and InterviewBit; Vijoy Basu, Senior Vice President and Global Head of Enterprise Data & AI, Firstsource; Tarun Bisht, VP, Technology, MPL; and Parimal Bajpai, VP of Engineering, InMobi.  Subhashis Sen, Cloud Solutions Director (JAPAC) – Digital Native, Oracle also participated in the discussion.

The roadblocks to scaling 

The roundtable focused on key scaling challenges startups face, from unexpected traffic spikes to bottlenecks that throw off a system’s performance. More than a few speakers discussed cost challenges, such as over-provisioning and network changes. In the fintech arena, speakers focused on the challenges of educating the public on scams and the high cost of acquiring new customers in niche markets. Variability in traffic patterns was also discussed, with speakers sharing how hard it was to predict scale.

AI and data challenges also played a large role in the conversation. Given the large volumes of data generated today, organisations find it difficult to mine the right insights. Speakers stressed the importance of bringing AI closer to data, bringing AI techniques closer to the source, for faster insights and more efficient workflows. 

The price of scaling

It is a tricky balance to maintain costs while scaling operations. Many speakers highlighted costs as one of the primary concerns when it came to scaling their cloud infrastructure. A solution that came up is serverless architecture, which makes it easy to process large volumes of data with efficiency, at a higher speed, at a cost that is 5X cheaper thanks to a pay-per-use model. Startups that are planning to scale should keep in mind that choosing the right architecture or a different tech stack can make a significant difference to costs. The discussion also spotlighted the importance of staying agile.

Oracle shared that the intent of setting up Oracle Cloud Infrastructure (OCI) was to provide both transparency and cost benefits, based purely on technical architecture. This goal was manifested through different approaches. When it came to the challenge of overprovisioning compute, using T-shirt sizing (a project estimation and capacity planning tool) could result in allocating more resources than necessary. In response to this, Oracle chose to not include T-shirt sizing, but opted for a more flexible approach, where clients can select the resources (for instance, RAM) and increase their needs as required.

Oracle also focused on high network costs. The company has been working on making reduced data charges a major benefit of its cloud services. OCI offers low networking prices that empower enterprises to move large masses of data, including services that consume a high volume of bandwidth. When it comes to AI issues, Oracle has dealt with the problem by building AI training features in the database, MySQL, and other database offerings. These offerings were created to strategically help lower costs by making technological and architectural changes.

A matter of latency

As a startup begins to scale, latency becomes an increasingly important issue to deal with. Companies can opt to scale vertically or horizontally, either adding more power to the existing server or more hardware to increase the computational power of the system overall. However, speakers pointed out that when it comes to databases, scaling with available database solutions can be difficult. Startups that choose to scale databases vertically often discovered that this came with increased latency issues. The proposed solution? Building in-memory databases to allow for faster access to data, which are particularly useful when it comes to real-time responses, high throughput and low latency. 

The complexity of compliance and security

With the Digital Personal Data Protection Act (DPDP Act) on the horizon, startups looking to expand can run into challenges navigating a slew of national and international regulations.

Speakers described the process as a grind, where companies need to stay cognizant of all current and upcoming regulations that apply to different territories before determining which country can get its own database. 

Participants also discussed how essential data security is in this day and age. Solutions like isolating production data to a small set of employees were discussed, along with techniques such as data obfuscation, which alters sensitive data making it less useful to unauthorized users. 

A unique challenge that was highlighted was the replication of data points across different microservices. Speakers pointed out how companies were struggling to encrypt or obfuscate data points in just one place, without it becoming a massive and costly endeavour. 

Panellists from the banking sector discussed their own challenges with compliance, sharing that in addition to acts like GDPR or the upcoming DPDP in India, they had to remain cognizant of banking regulations in each region, and then decide on data security techniques that aligned best with these territories. 

The discussion covered several significant topics, including monitoring cloud infrastructure, managing the cost of AI-oriented infrastructure, and the need for amortised cost calculations for experimental AI-driven projects.